Because the presumption is that the rats see the maze as the experimenter sees it, a featureless expanse with barriers. The fact recognized by the cited experimenter is that the rats are aware of various markers not evident to the human observers, and are able to recognize their location within a maze without “learning” the sequence of turns which brought them there. In other words, the rats “cheat”. Surprise! Surprise!
That doesn’t sound to me like the rats are “cheating.” It sounds only like they are using other than visual clues, which the researchers had assumed they were using. They still must navigate through the turns, and if there is a wall where there once was a passage, they must still find another way around. Even with the floor remaining the same, the rats have to learn the layout with each new experiment.